k__

11 jam yang laludiscuss

Yeah, DeepSeek V4 (Flash and Pro) is below $0.004 for 1M cache hits.

Even with usage based billing I'm below $1 writing code all day.

k__

kemarin duludiscuss

[dead]

k__

6 hari yang laludiscuss

"Europe's rapid warming is partly the result of [...] a drop in the number of tiny polluting particles in the air. This means that less of the Sun's energy is reflected back into space, leaving more energy to heat the Earth's surface."

Hmmm...

k__

7 hari yang laludiscuss

I tried DS4 Pro yesterday for one particular task that Flash struggled with.

That one Pro task cost $1.50 and the whole day of using Flash cost me $5.

k__

9 hari yang laludiscuss

Thanks!

I'll try that.

k__

9 hari yang laludiscuss

Yeah, v4 flash is dirt cheap, but it's running in circles quite often.

Might very well be that a better model is cheaper if it gets things right the first try.

Maybe I should route to a better model when v4flash hasn't solved after a specific number of tokens.

k__

9 hari yang laludiscuss

Nice.

I paid $6 yesterday for DeepSeek V4 Flash on OpenRouter. That's like $120 dollar for a month, and it's not even a good model.

k__

12 hari yang laludiscuss

Llama3.1 instruct seems to be doing okay on that page, mostly because it's dirt cheap.

k__

12 hari yang laludiscuss

So, their incentive is to promote AI music, since they don't have to pay royalties for them.

k__

23 hari yang laludiscuss

Did Nim 3 go all-in on compile time memory management?

k__

23 hari yang laludiscuss

I'm curious about Denuvo's opinion on that.

k__

23 hari yang laludiscuss

Maybe, you could pipe it through T5 or something.

k__

24 hari yang laludiscuss

Training DeepSeek was magnitudes cheaper than training the SOTA models it relied upon.

In theory, other countries should be able to replicate that effort and improve it.

k__

24 hari yang laludiscuss

As I understood it, they were only a US thing anyway. Because of healthcare issues etc.

k__

25 hari yang laludiscuss

I tried some smaller Gemma4 and Qwen3.6 quants on my MBA with M5/16GB and had like 20-60 tokens per second. At 60 it felt pretty okay and that hardware is on the lower end.

I'd assume a Mac with 32-64GB memory would get some reasonable results.

k__

25 hari yang laludiscuss

I thought the joke was that people aren't paying enough money.

k__

27 hari yang laludiscuss

100K seems quite much.

I had the impression, models would get inconsistent after just 3000 words.

k__

bulan laludiscuss

Why doesn't Mistral distill?

k__

bulan laludiscuss

I tried it with embedded programming, and failed miserably.